jaychempan/coding-with-beat/cwb-profile

coding-with-beat — Music Profile & Listening Report

You have access to the generate_profile MCP tool. Use it to analyse the user's listening history and generate personalised reports and recommendations.

When to activate

• 听歌报告 · 音乐画像 · 本周报告 · 本月报告 · 年度报告 · 日报告
• music profile · listening report · music report · my music taste
• 分析我的听歌 · 我最近在听什么 · 推荐基于历史
• history profile · what have I been listening to
• 给我推荐基于历史 · 根据我的喜好推荐

Dispatch logic

1. Detect period from the user's message:

• 今天 / 日 / today / daily → "daily"
• 本周 / 这周 / week / weekly → "weekly" (default)
• 本月 / 这个月 / month / monthly → "monthly"
• 今年 / 年度 / year / yearly → "yearly"
• No mention → default "weekly"

1. Extract context (optional scene/mood words the user mentions):

• 写代码 / coding / 跑步 / 放松 / 通勤 etc. → pass as context
• If no scene mentioned → context = ""

1. Call the tool:

`` generate_profile(period=<detected>, context=<extracted>) ``

1. Display the returned report in full.

1. After the report, present the recommendation queries and ask:

> "要播放推荐吗？说"播放推荐1"或直接告诉我你想听哪个方向。"

1. If user agrees to play:

• Extract the query strings from the report's recommendation section
• Call smart_search(queries=[<those queries>])
• Show results and let user pick by number with play_number(n)

Example exchanges

User: 帮我生成本周的听歌报告 → generate_profile(period="weekly", context="")

User: 我最近都在听什么，帮我分析一下，适合写代码的 → generate_profile(period="weekly", context="写代码")

User: 给我出一份年度音乐画像 → generate_profile(period="yearly", context="")

User: 今天听了什么 → generate_profile(period="daily", context="")

Error handling

If the tool returns "不足 5 首", tell the user:

"还没有足够的听歌记录来生成画像，先多听几首再来吧 🎵"

Do not retry automatically.